The Phlebographic Diagnosis of Deep Vein Thombosis — A Study in Observer Variation

Abstract

Three radiologists with varying experience, on two separate occasions, independently reviewed 161 lower limb phlebograms in 100 patients suspected of having deep vein thrombosis. A standard technique was used. Only two interpretations were made: either that thrombus was present or that it was not. To exclude agreement by chance, data was analysed both by calculation of percentage agreement and using the Kappa statistic. Intraobserver variation was not affected by re-examination of the same phlebogram by the same observer, suggesting that review is not necessary. In general the levels of agreement were similar between the three observers irrespective of their experience. Only in the ilio–femoral segment was there any significant disagreement between the most and the least experienced radiologist, probably because of less than ideal opacification. However, when the leg as a whole was considered there was similar level of agreement between the three observers. When chance agreement was not exluded, observers disagreed about the presence or absence of thrombus at some site in the limb in about 10% of examinations. We conclude from this study that the possibility of observer error in interpretation of a phlebogram should always be considered when managing patients with suspected deep vein thrombosis or pulmonary embolism.